Re: Conditional Requests to resolve semaphore and confidentiality concerns

On torsdag 16. januar 2020 23:38:17 CET james anderson wrote:
> > Admittedly, I do not understand your objection very well. In particular,
> > I'm not sure what you mean by meta-state, and therefore, I'm not sure
> > how to translate that requirement to the Web world.
> 
> in this case, the meta-state is whether the store is in a state in which it
> is to be permitted that a given process modify it.

Right, OK. What I'm trying to understand is in what sense you think about 
integrity. 

Perhaps it is worth entertaining this analogy: The Web permits the dreaded 
404, as it appeared (they way I understand the history) that it was terribly 
difficult to scale the Web if it was insisted on the strong integrity 
requirements of e.g. Xanadu. Yet, Tim insists that great effort should put 
into the integrity of the link, even though it isn't enforced. The result is 
that we live with dangling links, but we also have an actual global 
information system that works. 

I would like to further understand the debate we're having, because I do not 
quite understand what we are discussing at this point. Is there something 
that would obviously blow up so that it isn't usable to anyone with this 
approach?


> modulo clients which "possibly lie about their state”, that is the base
> use-case for a crdt.

Yes, we are fully aware of that, but we are also seeking to enable developers 
who are not prepared to go there to develop applications.

So, say that N and S are smaller, but not small enough to make full-document 
locks or etags feasible, is there a niche that could be occupied there?

Kjetil

Received on Friday, 17 January 2020 00:23:37 UTC